Merge branch 'master' into ci/gha-python

Author: Borda

docs/source/datamodules.rst

@@ -268,6 +268,7 @@ Use this method to generate the val dataloader.  Usually you just wrap the datas
         def val_dataloader(self):
             return DataLoader(self.mnist_val, batch_size=64)
 
+.. _datamodule-test-dataloader-label:
 
 test_dataloader
 ^^^^^^^^^^^^^^^

docs/source/optimizers.rst

@@ -201,12 +201,12 @@ For example, here step optimizer A every 2 batches and optimizer B every 4 batch
     # Alternating schedule for optimizer steps (ie: GANs)
     def optimizer_step(self, current_epoch, batch_nb, optimizer, optimizer_idx, closure, on_tpu=False, using_native_amp=False, using_lbfgs=False):
         # update generator opt every 2 steps
-        if optimizer_i == 0:
+        if optimizer_idx == 0:
             if batch_nb % 2 == 0 :
                optimizer.step(closure=closure)
 
         # update discriminator opt every 4 steps
-        if optimizer_i == 1:
+        if optimizer_idx == 1:
             if batch_nb % 4 == 0 :
                optimizer.step(closure=closure)
 
@@ -220,11 +220,11 @@ For example, here step optimizer A every 2 batches and optimizer B every 4 batch
     # Alternating schedule for optimizer steps (ie: GANs)
     def optimizer_step(self, current_epoch, batch_nb, optimizer, optimizer_idx, closure, on_tpu=False, using_native_amp=False, using_lbfgs=False):
         # update generator opt every 2 steps
-        if optimizer_i == 0:
+        if optimizer_idx == 0:
             optimizer.step(closure=closure, make_optimizer_step=(batch_nb % 2) == 0)
 
         # update discriminator opt every 4 steps
-        if optimizer_i == 1:
+        if optimizer_idx == 1:
             optimizer.step(closure=closure, make_optimizer_step=(batch_nb % 4) == 0)
 
 Here we add a learning-rate warm up

docs/source/test_set.rst

@@ -3,6 +3,10 @@
 Test set
 ========
 Lightning forces the user to run the test set separately to make sure it isn't evaluated by mistake.
+Testing is performed using the ``trainer`` object's ``.test()`` method.
+
+.. automethod:: pytorch_lightning.trainer.Trainer.test
+    :noindex:
 
 ----------
 
@@ -82,4 +86,22 @@ is not available at the time your model was declared.
     trainer.test(test_dataloaders=test)
 
 You can either pass in a single dataloader or a list of them. This optional named
-parameter can be used in conjunction with any of the above use cases.
+parameter can be used in conjunction with any of the above use cases. Additionally,
+you can also pass in an :ref:`datamodules` that have overridden the 
+:ref:`datamodule-test-dataloader-label` method.
+
+.. code-block:: python
+
+    class MyDataModule(pl.LightningDataModule):
+        ...
+        def test_dataloader(self):
+            return DataLoader(...)
+
+    # setup your datamodule
+    dm = MyDataModule(...)
+
+    # test (pass in datamodule)
+    trainer.test(datamodule=dm)
+    
+
+

pytorch_lightning/metrics/metric.py

@@ -94,7 +94,8 @@ def add_state(
                 reset to this value when ``self.reset()`` is called.
             dist_reduce_fx (Optional): Function to reduce state accross mutliple processes in distributed mode.
                 If value is ``"sum"``, ``"mean"``, or ``"cat"``, we will use ``torch.sum``, ``torch.mean``,
-                and ``torch.cat`` respectively, each with argument ``dim=0``. The user can also pass a custom
+                and ``torch.cat`` respectively, each with argument ``dim=0``. Note that the ``"cat"`` reduction
+                only makes sense if the state is a list, and not a tensor. The user can also pass a custom
                 function in this parameter.
             persistent (Optional): whether the state will be saved as part of the modules ``state_dict``.
                 Default is ``False``.
@@ -244,7 +245,7 @@ def reset(self):
         """
         for attr, default in self._defaults.items():
             current_val = getattr(self, attr)
-            if isinstance(current_val, torch.Tensor):
+            if isinstance(default, torch.Tensor):
                 setattr(self, attr, deepcopy(default).to(current_val.device))
             else:
                 setattr(self, attr, deepcopy(default))

tests/metrics/test_metric.py

@@ -26,6 +26,20 @@ def compute(self):
         pass
 
 
+class DummyList(Metric):
+    name = "DummyList"
+
+    def __init__(self):
+        super().__init__()
+        self.add_state("x", list(), dist_reduce_fx=None)
+
+    def update(self):
+        pass
+
+    def compute(self):
+        pass
+
+
 def test_inherit():
     a = Dummy()
 
@@ -77,12 +91,21 @@ def test_reset():
     class A(Dummy):
         pass
 
+    class B(DummyList):
+        pass
+
     a = A()
     assert a.x == 0
     a.x = torch.tensor(5)
     a.reset()
     assert a.x == 0
 
+    b = B()
+    assert isinstance(b.x, list) and len(b.x) == 0
+    b.x = torch.tensor(5)
+    b.reset()
+    assert isinstance(b.x, list) and len(b.x) == 0
+
 
 def test_update():
     class A(Dummy):